Robotic Disassembly

One of the main ideas of my Ph.D research is how to use a robot to try to take apart a laptop even if the robot has never seen that model of laptop before. The robotic setup that I am using is a 3 degree of freedom robot that was donated to URI in 2005. It was used by a graduate student before me to take out batteries in calculators. One problem with this robot is that most of its computers and electronics were outdated and proprietary so a Galil Motor Controller was installed.



This robotic setup had to be modified. This included creating an enclosure and making sure the lighting was uniform, creating a workspace stage that allowed for laptops of different sizes to be places on it,and creating a PCB that interfaced with the Galil Motor board. The new enclosure allowed for control over the environmental lighting and had removable black cloth around the wooden frame. The redesigned stage allowed for a uniform background as well as an automated clamping system with linear actuators rather than a manual clamp. The black cloth allowed for the IR beam sent out by the Kinect sensor to be absorbed rather than being bounced back and blinding the sensor.







A screwdriver end effector tool was also created and had a lot of different components incorpated on it including an accelerometer to determine when a screw was fully loosened, a electromagnet to pull a screw out of the hole, and a force sensing resistor to determine when the screwdriver had made contact with a surface. Additionally, another PCB was manufactured and was able to fit on top of an Arduino microcontroller.



There are two webcams that are used by the robotic system. The first webcam has an overview of the entire workspace and is used to determine circles that occur in the workspace. After the rough coordinates of these circles are determined the webcam that is on the robot is used to center and precise coordinates are determined for the screwdriver. The screwdriver proceeds to each location to probe and determine if a screw is actually present. The first image below shows all of the different circles that are found using a circle transform algorithm and the second image shows the results of probing those locations with the screwdriver (green means a screw was found while the other colors are a screw was not found for various reasons).





Future work includes incorporating a cognitive system to remember the location of the screw holes on different models so the disassembly time on known models can be reduced.